Soft Patch Panel
 help / color / mirror / Atom feed
* [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid
@ 2020-04-19  7:02 bugzilla
  2020-04-21  9:24 ` bugzilla
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: bugzilla @ 2020-04-19  7:02 UTC (permalink / raw)
  To: spp

https://bugs.dpdk.org/show_bug.cgi?id=457

            Bug ID: 457
           Summary: rte_mempool_create is based on lcore socketid and not
                    the port socketid
           Product: SPP
           Version: unspecified
          Hardware: x86
                OS: Linux
            Status: UNCONFIRMED
          Severity: major
          Priority: Normal
         Component: main
          Assignee: yasufum.o@gmail.com
          Reporter: vipin.varghese@intel.com
                CC: spp@dpdk.org
  Target Milestone: ---

File: ```src/primary/init.c```
line: 52

current: 
```
pktmbuf_pool = rte_mempool_create(PKTMBUF_POOL_NAME, num_mbufs,
                        MBUF_SIZE, MBUF_CACHE_SIZE,
                        sizeof(struct rte_pktmbuf_pool_private),
                        rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
                        rte_socket_id(), NO_FLAGS);
```

expected
```
/* using rte_eth_dev_socket_id (uint16_t port_id) 
   a. identify the sockets and pool to be used 
   b. number of ports in same pool
 */
pktmbuf_pool[0] = rte_mempool_create(PKTMBUF_POOL_NAME_0, num_mbufs_0,
                        MBUF_SIZE, MBUF_CACHE_SIZE,
                        sizeof(struct rte_pktmbuf_pool_private),
                        rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
                        0, NO_FLAGS);
```

-- 
You are receiving this mail because:
You are on the CC list for the bug.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid
  2020-04-19  7:02 [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid bugzilla
@ 2020-04-21  9:24 ` bugzilla
  2020-04-21 11:06 ` bugzilla
  2020-04-23  5:04 ` bugzilla
  2 siblings, 0 replies; 4+ messages in thread
From: bugzilla @ 2020-04-21  9:24 UTC (permalink / raw)
  To: spp

https://bugs.dpdk.org/show_bug.cgi?id=457

Hideyuki Yamashita (yamashita.hideyuki@ntt-tx.co.jp) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |yamashita.hideyuki@ntt-tx.c
                   |                            |o.jp

--- Comment #1 from Hideyuki Yamashita (yamashita.hideyuki@ntt-tx.co.jp) ---
Hello Vipin,

Thanks for your comments.
However I could not understand what is the real problem on the current code.
I mean primary process can be run any lcore id which can be specified in
primary
process startup parameter (with -l option).
With -l option, any lcore id can be specified. In other words, either of NUMA
node
(e.g. numa0 numa1) can be specified.
Original coding (calling rte_socket_id()) is considering above situation.
However your "Expected" codeing specifies "soeket id = 0".

-- 
You are receiving this mail because:
You are on the CC list for the bug.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid
  2020-04-19  7:02 [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid bugzilla
  2020-04-21  9:24 ` bugzilla
@ 2020-04-21 11:06 ` bugzilla
  2020-04-23  5:04 ` bugzilla
  2 siblings, 0 replies; 4+ messages in thread
From: bugzilla @ 2020-04-21 11:06 UTC (permalink / raw)
  To: spp

https://bugs.dpdk.org/show_bug.cgi?id=457

--- Comment #2 from Vipin Varghese (vipin.varghese@intel.com) ---
Hello Hideyuki,

I am not clear with your comment, are you sharing

```
even if the Interface is NUMA-1, since you have given core mask from NUMA-0
there is no performance impact?

or 

there is a performance impact on incorrect NUMA core mask and NUMA NIC. In
current software implementation, since the code is absent it is desired `not to
start SPP-Primary with multi NUMA ports. If there are multiple NUMA NIC, one
has to have multiple SPP-Primary per NUMA to get over the issue`.
```

Can you show me where this is documented?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid
  2020-04-19  7:02 [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid bugzilla
  2020-04-21  9:24 ` bugzilla
  2020-04-21 11:06 ` bugzilla
@ 2020-04-23  5:04 ` bugzilla
  2 siblings, 0 replies; 4+ messages in thread
From: bugzilla @ 2020-04-23  5:04 UTC (permalink / raw)
  To: spp

https://bugs.dpdk.org/show_bug.cgi?id=457

masahiro nemoto (masahiro.nemoto.es@s1.ntt-tx.co.jp) changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |masahiro.nemoto.es@s1.ntt-t
                   |                            |x.co.jp

--- Comment #3 from masahiro nemoto (masahiro.nemoto.es@s1.ntt-tx.co.jp) ---
Hello Vipin,

1. Is our understanding correct?
Your concern:   Under NUMA node configuration,  it is possible that Primary is
started on NUMA-0 while NIC is under NUMA-1.  No instruction, warning in SPP
document to restrict such case.

2. Our opinion
We would prefer to add some text which restrict above use case at least clearly
states performanc degladation.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2020-04-23  5:04 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-04-19  7:02 [spp] [Bug 457] rte_mempool_create is based on lcore socketid and not the port socketid bugzilla
2020-04-21  9:24 ` bugzilla
2020-04-21 11:06 ` bugzilla
2020-04-23  5:04 ` bugzilla

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).